Baseline patient-reported symptom networks and early risk stratification for six-month symptom burden in newly...

To characterize the baseline patient-reported symptom network in newly diagnosed glioma and to develop a baseline-only model for predicting high symptom burden at 6 months, with the aim of informing earlier symptom surveillance and risk-stratified supportive care.

In a prospective cohort of 142 adults with histologically confirmed glioma, 15 patient-reported symptoms were rated at baseline using 0-10 numeric rating scales. Baseline symptom networks were estimated using Gaussian graphical modeling with graphical LASSO regularization and EBIC model selection. Edge accuracy and centrality stability were evaluated using nonparametric bootstrap and case-dropping procedures. High symptom burden at 6 months was defined using an upper-quartile rule based on the cohort distribution of the 6-month total symptom score. A random forest model incorporating baseline symptoms, clinical variables, and psychosocial factors was trained using a stratified split and internally validated in an independent test set.

The baseline network was moderately dense (38 edges; density = 0.36) and showed prominent cross-domain interdependencies. Speech difficulty and headache had the highest strength centrality, whereas betweenness and closeness showed lower stability. The random forest model achieved an AUC of 0.947 (95% CI, 0.888-1.000), and the leading predictors were baseline anxiety, depression, memory impairment, pain, fatigue, and age.

Symptom burden in newly diagnosed glioma appears to arise from an interconnected patient-reported symptom pattern in which emotional, cognitive, physical, and neurological symptoms interact closely. Baseline multidimensional assessment and early risk stratification may help identify patients who warrant closer symptom surveillance and supportive care during the initial treatment course. These findings provide a basis for further external validation.